The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external content from blog files to generate narration scripts. An attacker-controlled blog post could attempt to inject malicious instructions into the narration preparation phase.
Ingestion points: Reading content from user-specified blog files in SKILL.md (Step 1) and read_text_file in scripts/generate_audio.py.
Boundary markers: None identified between external content and instructions.
Capability inventory: The skill has the ability to write files to the local disk and execute subprocesses (ffmpeg).
Sanitization: The skill relies on the LLM to strip markdown and clean text for narration, but no programmatic sanitization is applied before interpolation.
[COMMAND_EXECUTION]: The skill uses subprocess.run for legitimate administrative and processing tasks.
scripts/setup_environment.py executes pip to install dependencies.
scripts/generate_audio.py executes ffmpeg to convert WAV files to MP3.
scripts/run.py executes internal skill scripts within a managed virtual environment.
[EXTERNAL_DOWNLOADS]: The skill fetches necessary Python packages from the official registry during the setup phase. It mitigates supply chain risks by using a lock file (requirements.lock) with SHA-256 hash verification.

blog-audio